Project Summary Social learning?the ability to learn from others?is essential for adaptive behavior. Humans not only receive social information, but also actively interpret it in light of the unobservable mental states of the person providing it (e.g., their knowledge, goals, and preferences). This ability to reason about others? mental states, known as Theory of Mind (ToM), allows humans to capitalize on sparse, imperfect social information. The central objective of this proposal is to characterize the neural computations that enable humans to use ToM to guide decision-making. Specifically, this proposal integrates computational models of ToM with fMRI to ask how the outputs of mental state inferences are represented in neural systems that support ToM and value-based decision-making (Aims 1 & 2) and how these systems interact to support flexible behavior (Aim 3). My dissertation work provides a computational account of how adults use mental state inferences to make decisions that benefit the self (Aim 1.1) and others (Aim 1.2). In recently published work, Aim 1.1 found that adults adjust their use of advice based on the advisor?s knowledge, intent, and strategy; participants? behavior was best described by a Bayesian ToM model that infers the value of a hidden option that is only known to the advisor. Aim 1.2 found that adults can generalize others? preferences to a set of novel options; these results suggest that adults? representations of others? preferences are grounded in abstract, generalizable features. This work has provided extensive training in computational models of cognition. Work proposed during the F99 phase will use fMRI to test hypotheses about the neural implementation of these computations. Building on Aim 1.1, Aim 2.1 tests whether neural value signals track the value of hidden options that must be inferred from advice. Building on Aim 1.2, Aim 2.2 tests whether the representational structure of patterns of activity shifts during choice, based on whether participants are considering their own or their partner?s preferences and on which features the person for whom they are choosing values. This work will provide extensive training in model-based fMRI and will capitalize on the tools and training available in my graduate institution to bring this work to a high standard of transparency and computational reproducibility. Work proposed in the F00 phase will examine how humans allocate cognitive resources to mental state inference. This work will test the overarching hypothesis that humans choose between mental state inference and other, simpler strategies by balancing mental effort with expected information gain. This work will build on my expertise in model-based fMRI by providing additional training in connectivity-based fMRI methods, information theory, and resource-rational cognitive models. In summary, the proposed work will yield new insights into computational and neural underpinnings of decision-making in social contexts and will provide the training needed for me to become an independent researcher at the forefront of social cognitive neuroscience.